Dave's power supply mods.

[firewalker]

Post here you mods/ideas for Dave's design.

Alexander.

[Blue]

Higher output voltage. E.g. 18V out...

[don.r]

Higher output voltage. E.g. 18V out...

Higher outputs would entail losing the resolution from the MCU. You may need to step up to 16bit converters. I always like extra fixed outputs like +/- 5V and +/- 12V along with the variable. Maybe with a fixed or switch selectable current limit. Easily done with 78XX/79XX parts.

[PeteInTexas]

Voltage and current presets.  I find that I actually only use a handful of voltages that correspond with common battery combinations- 1.5, 3.0, 6.0, 9.0 and so it is annoying to have to keep turning a 10 turn pot.  With the 12-bit DAC, I can imagine it will be similarly annoying with the rotary encoder.

It would really be great to just go to the voltage you want.  A 16-key pad mod would work using the 4 encoder pins plus the 4 pins from the 4 switches.  That still leaves too much twiddling for my tastes.

Unfortunately, the 4 encoders are not going into ADC pins so a resistor matrix cannot be used for preset switches.  They still can give 16 values for the software so it is possible to portion that out between voltage and current presets.  If the 4 switches can be similarly put into only two pins, one of the saved pins can be used to toggle between the presets and the rotary encoders.  The best of both worlds!

[PeteInTexas]

If preset switches ain't your bag, you can multiplex the four switches into 2 pins and use the 2 saved pins to toggle between 'coarse' and 'fine' encoder steps for voltage and current respectively.

[Bored@Work]

If preset switches ain't your bag, you can multiplex the four switches into 2 pins and use the 2 saved pins to toggle between 'coarse' and 'fine' encoder steps for voltage and current respectively.

Ehm, one would do that by applying a dynamic-speed encoder profile. If you turn the encoder fast you do larger voltage or current steps (coarse). If you turn the encoder slowly you do fine steps.

[ndictu]

Ehm, one would do that by applying a dynamic-speed encoder profile. If you turn the encoder fast you do larger voltage or current steps (coarse). If you turn the encoder slowly you do fine steps.

That's a great idea, and you just need a bit more code to do that so it'll be easy.

I kind of lost track of the discussion in the main thread, anyone knows what Dave plans on using for the code? It would be best on github/other VCS, so that everyone can easily submit their modifications and they will get merged to the "official" version.

Also, few ideas:
- use the serial link to connect multiple supplies together, using one as a master control for the others
- with isolated USB link you can do all kinds of stuff: log all the values, measure power consumption, automatically sweep over the voltage range and take measurements after some settle period etc...
- using a bigger uC and duplicating most of the supply circuitry you can transform it into a dual output supply. But I would probably rather build two of these.

[Zad]

Dave has said that there is one thing he isn't telling us about the design yet. But that could be anything

One possibility is that he is using an I2C expander IC, and using that to drive a standard LCD and key switches. I think a lower power version built into an Arduino shield could potentially be a good seller via Sparkfun, Adafruit etc.

Fancy things like battery charging, preset profiles and so on are all just software and can be left to the community. Such is the benefit of Open Source. Dave mentioned in the video the potential for an opto-isolated RS232 control and logging, which is easy to connect to USB via a cheap external interface.

[Bored@Work]

Dave has said that there is one thing he isn't telling us about the design yet. But that could be anything

I had the impression it is more about the enclosure. And I seem to remember that in some older video he was thinking about a certain kind of enclosure for some projects.

[PeteInTexas]

If preset switches ain't your bag, you can multiplex the four switches into 2 pins and use the 2 saved pins to toggle between 'coarse' and 'fine' encoder steps for voltage and current respectively.

Ehm, one would do that by applying a dynamic-speed encoder profile. If you turn the encoder fast you do larger voltage or current steps (coarse). If you turn the encoder slowly you do fine steps.

This relies too much on ones coffee in take to finally get to the desired value.  I was thinking in the coarse mode each step cycles through the desired set of presets.  So for my 4 voltages, 3 twiddles at the most- 1 to flip to coarse mode, 2 to get to the desired voltage (you can go backwards).

[PeteInTexas]

Ehm, one would do that by applying a dynamic-speed encoder profile. If you turn the encoder fast you do larger voltage or current steps (coarse). If you turn the encoder slowly you do fine steps.

That's a great idea, and you just need a bit more code to do that so it'll be easy.

I kind of lost track of the discussion in the main thread, anyone knows what Dave plans on using for the code? It would be best on github/other VCS, so that everyone can easily submit their modifications and they will get merged to the "official" version.

Also, few ideas:
- use the serial link to connect multiple supplies together, using one as a master control for the others
- with isolated USB link you can do all kinds of stuff: log all the values, measure power consumption, automatically sweep over the voltage range and take measurements after some settle period etc...
- using a bigger uC and duplicating most of the supply circuitry you can transform it into a dual output supply. But I would probably rather build two of these.

Good ideas but I make a distinction between a mod and a rework.  You are asking for a rework.

[Bored@Work]

This relies too much on ones coffee in take to finally get to the desired value.

Are you able to operate a computer mouse under the influence of coffee? If yes, you should be able to operate this, too. Because dynamic profiles are the norm for mouse operations since ages.

[PeteInTexas]

This relies too much on ones coffee in take to finally get to the desired value.

Are you able to operate a computer mouse under the influence of coffee? If yes, you should be able to operate this, too. Because dynamic profiles are the norm for mouse operations since ages.

It took me a long time to perfect my mouse setting. 

[Short Circuit]

..
Ehm, one would do that by applying a dynamic-speed encoder profile. If you turn the encoder fast you do larger voltage or current steps (coarse). If you turn the encoder slowly you do fine steps.

This relies too much on ones coffee in take to finally get to the desired value.  I was thinking in the coarse mode each step cycles through the desired set of presets.  So for my 4 voltages, 3 twiddles at the most- 1 to flip to coarse mode, 2 to get to the desired voltage (you can go backwards).

Nonsense, if implemented correctly, that works marvelous. Course/fine switches on the other hand are a pain.
Always in the wrong mode, so twist, oops backtwist, toggle, twist again, toggle, fine twist.
I have a PSU that is operated like this and it sucks.

[PeteInTexas]

This relies too much on ones coffee in take to finally get to the desired value.  I was thinking in the coarse mode each step cycles through the desired set of presets.  So for my 4 voltages, 3 twiddles at the most- 1 to flip to coarse mode, 2 to get to the desired voltage (you can go backwards).

Nonsense, if implemented correctly, that works marvelous. Course/fine switches on the other hand are a pain.
Always in the wrong mode, so twist, oops backtwist, toggle, twist again, toggle, fine twist.
I have a PSU that is operated like this and it sucks.

I'm guessing you harboring similar disdain for oscilloscopes?

[Rerouter]

For higher Voltage
(and this is only as a higher voltage modification 30.72V)
2 x http://search.digikey.com/us/en/products/MC33172N/497-7409-5-ND/1038900 op amps are pin compatible with u3 and u12

a few gain resistor changes;

2K in parrellel with R29
2K in parrellel with R31 (for a gain of 15)

10K in parrellel with R35
82K to replace R34 (for a maximum input voltage reading of approx 35.6V in, yeah not a pretty number, still working)

and for the mosfet, like a 82K resistor tieing its gate to ground, to divide input voltage by 2 (about 18V max), so its safe from 0 to 30.72V out,

For a keypad / more free pins
Hook up the keypad / existing buttons to a resistive ladder / matrix and in software make use of the adc function of the pin some of the buttons are currently on

For higher current
being how the LT3080 only requires 20mohm between regulators one could drill additional holes into the heatsink and mount regulators hooked up in parrellel, even with mod wire it would roughly be enough,

[shebu18]

This one would be nice for a dual powersupply. A LT3080 or a L338 that supports negative voltage and a bigger micro, and one more from everythgin plus some other parts.


The biggest problem vould be the box for the PSU.

[naimis]

mains inputs? Or am I being altogether too obvious? 

[shebu18]

After rectifing, filtering and filtering again. I think you get arround 17V, even 14 will be ok.

[caroper]

What I have in mind is to use 2 of Dave's Boards (or something based on his design) but keep them isolated and floating so that they could be connected in Series for grater voltage or a -V_0_V+ Supply or paralleled for a larger current output. They would have to be in close tolerance but I think Dave's design has taken that into account.  I would leave the MCU on the board for the sake of reading the Encoder and driving isolated communications via SPI or TTL Serial, but not worry to much about adding any advanced functions to it. Possibly not even a display at this stage as you could use a DMM to set the Current limit and Measure the Output Voltage.

Later I would add a separate control board, with a larger processor, to take over the display and control input functions, with software to add tracking and other optional capability's such as computer control, voltage sweeps, stepped patterns etc. for bench testing of devices.

As part of the CPU board I would include a relay control to automatically configure the outputs into Independent / Serial tracking / Parallel Tracking configurations. Because it has to be isolated from the 2 power channels and independently powered it could also provide fixed outputs of 3V3. 5V and possibly 9/12V if it is driving relays. thus providing additional common supply voltages and freeing up the 2 main channels for more productive tasks.


If i were to change the current design at all it would be to simplify it rather than add features.

Even with only 10bit resolution on ADC and DAC you could control 0->1024mA in steps of 1mA and 0->20.48V in steps of 20mV, which I think would be more than adequate for most hobby bench supplies and indeed exceed the measurement capability's of most hobbyist DMM's, so I would be inclined to use PWM and on chip ADC of the MCU rather than increase the cost and complexity with the external devices. For the same reason I would argue that the inclusion of the uCurrent is also an unnecessary expense. If I were designing something that had low voltage and current requirements, such as MCU devices that draw down to the uA range in sleep mode and nA normally, I would not be wasting 10's of watts of power by testing them on a bench PSU like this one. For those applications a precision PSU of 0->6V @ 0->10mA with very fine resolution would be a good investment.

[EEVblog]

Even with only 10bit resolution on ADC and DAC you could control 0->1024mA in steps of 1mA and 0->20.48V in steps of 20mV, which I think would be more than adequate for most hobby bench supplies and indeed exceed the measurement capability's of most hobbyist DMM's, so I would be inclined to use PWM and on chip ADC of the MCU rather than increase the cost and complexity with the external devices. For the same reason I would argue that the inclusion of the uCurrent is also an unnecessary expense.

They don't actually add a huge lot to the overall cost when you add it all up. It's not like the component cost suddenly halves by taking out those components.

If I were designing something that had low voltage and current requirements, such as MCU devices that draw down to the uA range in sleep mode and nA normally, I would not be wasting 10's of watts of power by testing them on a bench PSU like this one.

This is a linear supply, it uses the same current as the device under test + some the power for the circuitry.
So a circuit taking say 1mA at 3.3V only takes 1mA from the input voltage as well. In this case the PSU circuit power dominates. I don't know what that is yet, but maybe another 20mA or so.

For those applications a precision PSU of 0->6V @ 0->10mA with very fine resolution would be a good investment.[/size]

That's what this supply is. A small precision 0-10V 0-2A supply with fine resolution.
0-10mA capability only would be a waste of space IMO and very niche, 0-100mA would be better but still very limiting, 0-1A makes it a very usable supply for general circuits.

Dave.

[caroper]

I did actually intend 100mA the 10 was a typo, but point taken.  I was thinking of the voltage drop across the regulator as wasting power, but at only a few mA it is not dissipating much power at all.

I would still like to see 12->15 volts as the upper limit, but if they can be run in series it gives 20V for charging battery's etc and 10<-0->10 for driving op-amps in split rail mode so probably covers most bases.

My other problem with the external ADC/DAC is that you had to abandon through hole for them, not really the cost factor. But as you have said, it is for  precision with fine resolution so a valid compromise.
You may have to offer the KIT with those parts already mounted or you could lose a large potential market though.

EDIT:Spelling

[EEVblog]

My other problem with the external ADC/DAC is that you had to abandon through hole for them, not really the cost factor. But as you have said, it is for  precision with fine resolution so a valid compromise.
You may have to offer the KIT with those parts already mounted or you could lose a large potential market though.

The ADC and DAC are both DIP parts, available in 10bit and 12 bit versions.
The current sense amps are the ones that are SMD.
They are only SO8, pretty trivial to solder. If I was to get just them assembled, then I might as well get the whole PCB assembled, as the cost difference wouldn't be much. Then you might as well go all SMD to lower the cost, and so on it goes...

Dave.

[IanB]

I really have no problem with SMD in a kit. It's not like soldering through hole components is actually easy. Sometimes through hole devices can be damned hard and annoying!

[EEVblog]

I really have no problem with SMD in a kit. It's not like soldering through hole components is actually easy. Sometimes through hole devices can be damned hard and annoying!

Yeah, but most people don't see it that way, SMD scares them.
And SMD parts are a real PITA to kit up for and handle and sort etc, and people lose them, or can't read which cap is what etc.
That is why I don't do the uCurrent as a kit any more.

Dave.

[PeterG]

Is there a need for a 'Basic' version of this kit, with just the board and parts?
Even a board without the micro but with marked pins for a micro to be connected not even a case. This would make it more flexible for those who want to make there own custom psu's. In my case, a Keypad and maybe a 4x20 rgb lcd all running of an Arduino Mega1280 board.

Just an idea.

Regards

[EEVblog]

Is there a need for a 'Basic' version of this kit, with just the board and parts?
Even a board without the micro but with marked pins for a micro to be connected not even a case. This would make it more flexible for those who want to make there own custom psu's. In my case, a Keypad and maybe a 4x20 rgb lcd all running of an Arduino Mega1280 board.

Yes, I expect some just to buy it for that.
Although the board is custom designed for my case, mounting holes mean it can be mounted in any bigger case on standoffs.

Dave.

[firewalker]

I made a quick lookup at DigiKey for another linear voltage regulator with down to zero capability and... Didn't found one!

I thought there would be more than LT308x.

Alexander.

[JimmyM]

I was thinking that if an additional opamp (call it Ux) were placed between the opamp U3A and the LT3080. The output of U3A could go into the non-inverting input of Ux and the inverting input of Ux was connected to a voltage divider at the very end of the output. The output of Ux would then be connected to the input of the LT3080 where U3A was connected. Any voltage drops caused by current sensing, etc would be compensated for. Am I wrong in how this would work? This could also give a kelvin sense capability (at least for the V+ wire)
Would this be worth looking into since the LT3080/3085 does not have an external voltage sense connection. Also, the LT3085 is capable of 3 Amps which would amplify voltage drops caused by any i resistance between the output of the LT3085 (or multiple LT3080s)

[firewalker]

I think you mean LT3083. LT3085 is 500 mA.

Alexander.

[JimmyM]

I think you mean LT3083. LT3085 is 500 mA.

Alexander.

yes. Right. Sorry.

[EEVblog]

I made a quick lookup at DigiKey for another linear voltage regulator with down to zero capability and... Didn't found one!

I thought there would be more than LT308x.

I think there are a couple of others from memory in some form, but that's what makes the LT308x special, unless you want to roll your own, or use negative supplies to compensate etc.

Dave.

[alm]

I'm quite sure that Dave wouldn't have picked the LT3080 if it didn't have a significant advantage over the common parts. An experienced desigs engineer like Dave will avoid specialist, singe-sourced parts if possible. If the circuit would have worked just as well with an LM317, this would be the superior choice, due to the availability it's much easier and cheaper to source, so much less likely to cause supply issues. The LM317 won't go down to 0V however without adding extra complexity, hence the exotic LT3080.

[siliconmix]

whay would you want to go down to zero ?

[IanB]

whay would you want to go down to zero ?

For instance if you wanted to check the low battery function with a single AA cell you would want to go down to 0.6 or 0.7 V. Since an LM317 bottoms out at 1.25 V it would be unsuitable for this application.

[siliconmix]

whay would you want to go down to zero ?

For instance if you wanted to check the low battery function with a single AA cell you would want to go down to 0.6 or 0.7 V. Since an LM317 bottoms out at 1.25 V it would be unsuitable for this application.

oops just watched the vid

I'm quite sure that Dave wouldn't have picked the LT3080 if it didn't have a significant advantage over the common parts. An experienced desigs engineer like Dave will avoid specialist, singe-sourced parts if possible. If the circuit would have worked just as well with an LM317, this would be the superior choice, due to the availability it's much easier and cheaper to source, so much less likely to cause supply issues. The LM317 won't go down to 0V however without adding extra complexity, hence the exotic LT3080.

[jpelczar]

Another idea after watching Makerbot episode: how about providing a connector for the ATX PSU ? There are lots for unused PSUs lying around. This would allow for cool feature: power on/off by tact switch

[mobbarley]

A bit bored today.. Do I win the title of the first mod?




PS: what have I missed?

[mobbarley]

eagerly awaiting part 8 with details of the pcb / case so that I know if this will fit..... 

[senso]

I have already created the necessary footprints for the max parts and for the lm3080 and copied the schematic to Eagle, but I'm using an atmega644p instead and the front panel with the rotary encoders, buttons and display will be included in a separated board.

[Rerouter]

Was just reviewing part 3 of the video and picked up on the fact that the constant current source only works down to 0.8V, but with daves external one, seemed to go down to 22mV,

it would seem with a fet with a 400mV threshold, and driven at the 2.048V reference would capable of handling down to the last 8 bits or 20mV, while it would sink 27mA under the highest voltage, i felt i would just raise the suggestion,

perhaps using another op amp, one could invert the gain of Vset and have it level out would be the solution to that,

also i realise there is little need to get that low, but at a guess atleast one of us must be interested in that capability,

[JimmyM]

I have already created the necessary footprints for the max parts and for the lm3080 and copied the schematic to Eagle, but I'm using an atmega644p instead and the front panel with the rotary encoders, buttons and display will be included in a separated board.

Keeeeep talking. I like what I'm hearing.

[george graves]

I hope there are some pre-set voltages.  Seems a waist to have a micro in there if it's not going to do the work for you.

Also - wouldn't mind a bigger screen, and big readable fonts.  Not that my eye sight is bad, it just makes it easier to read out of the corner of your eye, or at a quick glance.



Oh and a cool case - it's gotta have a cool looking case.  I wish there was a generic version of something like this available:



A keypad would be cool - but I'm not sure how useful it would be.


Or - I might make something out of wood and stain it - this guy used an old hard drive for the top and bottom.

[Rerouter]

the 4 buttons will be left to us to chose there purpose,

also are there any downsides, (oscillations, offsets, etc) from using mosfets to switch op amp gains,

as it would be nice, wanting higher voltage ranges, to switch down the gain as i fall below 10V and regain the original accuracy, 

[PA3BNX]

Hello Every Body out there,


I have a question for Dave and his new designed powersupply.

Will it get a so called foldback current mechanism in software.

I mean if I short the power supply for longer as lets say three seconds

Will it switch the Current down to zero mA and the Voltage also to zero Volts?

So that I have to reset it before the Voltage comes up again ?

[Rerouter]

once again, i would think that would be something you might have to go enable, just because some people will want foldback and some will want it to hold, also the device itself will foldback at 1.2A, in thermal protection mode,

[ejeffrey]

Will it get a so called foldback current mechanism in software.

I mean if I short the power supply for longer as lets say three seconds

Will it switch the Current down to zero mA and the Voltage also to zero Volts?

So that I have to reset it before the Voltage comes up again ?

I am sure you could implement that, but foldback current limiting is not normally wanted in a lab supply and what is normally meant by it is not actually what you have described.

The normal purpose of foldback current limiting is to protect a fixed voltage linear power supply in case of a short circuit.  The problem is that when a load exceeds the current limit, the voltage will start to drop, but this increases the power dissipation in the pass transistor.  In the ultimate limit, when you have short circuited the output, the pass transistor(s) are dissipating more than 100% of the supply rating.  Bipolar transistors also have second breakdown which further limits their safe operating area.  Foldback current limiting uses a separate analog feedback loop to reduce the current limit and keep within the transistor SOA when the output voltage is below the setpoint.  Generally once the short is removed the supply will recover naturally -- it doesn't require a hard reset.  Of course, if the short circuit is caused by a crowbar circuit, it won't reset until the power is removed.

A software defined limit with a 3 second delay is not nearly fast enough to provide protection for the majority of circuits that need foldback current limiting.  That might help if your thermal limit is an undersized heatsink, but more often you have to deal with the package juction->case rating or second breakdown.  These time constants are quite fast, 3 seconds is much too long.

In any case, by design a bench supply must not be susceptible to such overloads.  The supply has to be able to survive being programmed to the lowest possible voltage and the highest possible current.  The transistor doesn't care whether the set point is 10 volts and drops to 1 volt due to the current limit, or whether the setpoint is 1 volt.

In any case, the LT3080 has built in thermal protection.  If you do manage to overload it, it will shut off to protect itself.  However, this would be a sign of a design error, or use of the supply outside its design parameters (i.e., too high ambient temperature).

If you want to protect the device under test rather than the supply, that would normally be considered overcurrent protection, and then you would generally design it to shut off the supply until reset.  There are not that many situations where you need overvoltage / overcurrent protection, and again, usually 3 seconds is way to long -- if you are powering a very delicate transducer or something, normally you would want to shut off the supply within microseconds.

[Rerouter]

or in those cases, you would set the maximum current just above expected current of the device, so that it holds there, and will fall to near 0V if a bad short was made

[amspire]

I don't think foldback current limiting has any place in a lab power supply.

If you set the current to 1A limit, you want it to limit to 1A, and not an unknown current that you haven't specified.

In using a power supply, I am using it regularly as a current source, and also in testing a new circuit or a faulty circuit, I want to be able to set the voltage and slowly wind up the current limit. Last thing I want is foldback decreasing the current as I am trying to increase it.

Foldback has a place in some fixed output voltage supplies where the regulators cannot maintain full current at 0V, but I would never build a variable lab supply that cannot maintain full current at 0V.

Richard.

[SeanB]

I think it is pretty good as a low power PSU with a really accurate output voltage setting ability. If you wanted a higher rating then you would not need the resolution, and a higher current rating with lower resolution would be perfect. It is a precision supply, good enough to use as a 4 digit reference voltage, or a 3 digit current source. Would be good to build in a small case and have a companion ( same software and interfacing but coarser steps) lookalike that does 0-40V at 0 - 5A, possibly using a multitapped transformer to reduce dissipation in the regulator stage. You could even build them into a larger case if you made the precision one a floating unit, though them your interfaces would need to be opto isolated if you were going to have remote setting ability. a 2/4 port USB hub and 2 USB to RS 232 converters and a set of opto isolators would do that, then a single USB socket on the back will do. The extra USB ports could then be front panel mounted for convenience - extra points to connect a USB scope that is still isolated from the power supply rails.

[PA3BNX]

Hello EveryBody,


My goal for foldback was simply when something unattended was running
and going bad that it switches off..

So no smoke in the device under test while away ?

I assume  that the Power Supply it self can with stand
the shortage.

If I put a a load with a big capacitor on the output of the power supply
then it should not foldback.

That's what the 3 second delay should be for.

It would be nice if you could see at what time the error function did occur.


 

[Rerouter]

that would all be possible through rolling your own code for it, say mapping one button to foldback/monitor,

[Rerouter]

well, after playing around with the design to handle a higher voltage range (30.72 - 10mV), this is a section that i feel need a second set of eyes on,

the current source is replaced with an error amp fet, driven at ~1mV, and the capacitor was to prevent oscilations at low loads, though i feel it may be an incorrect value,

the main output gain is stepped up to 15, and by pulling the fet on the gain down, it swaps back to ~10.24V, although 100K's would be more accurate, its getting into the noise for the op amp i think, and may be wrong on it,

the actual op amp model can be ignored, i havent updated it since the rev A board was released, and will need to be a rail to rail if not already,

other than that, i flipped the voltage setting portion around to try and tidy it up, and hoping for some constructive input,

[Rerouter]

And now the compliment sense line(s)

the fet once again switches between a gain of 5, and of 15, but my primary concern is, with the input clamped through 20K, even worse case, with the fet left off, would the 1.2mA passed by the diode clamp actually interfere with the reference in any measurable level? (30V/25K*5K =6V, 24V/20K = 1.2mA)

the reason why i want to clamp to the reference is, the op amp feeding the adc can possibly pass more current than its internal clamps can handle, and so want to minimise the risk,

[amspire]

Rerouter,

You have some ideas that can work, but you might be over complicating things.

First though, if you use a single LT3080, then if you want to output to go to over 30V, then you will have to reduce the current to about 250mA. Alternatively, you parallel 4 LT3080's at the cost to a slightly degraded voltage regulation due resistors you have to add to the LT3080 outputs.

Which way were you thinking of going?

Now Dave had some very particular design restrictions, that meant he absolutely had to be able to run his design from a single supply. 30V+ output sounds like you have a transformer, and once you have a transformer, getting a low current negative supply only costs some cheap diodes and capacitors (even if the transformer is just a single winding) and perhaps a 79L05 regulator. It would be nuts if you can have a negative supply, to not use it. You get much better regulation even at 0V out, the current source fed from the negative rail can sink down to 0V without any problem, you can use easy-to-get LM317's instead of LT3080, since you can offset the control voltage by -1.25V. A rail to rail opamp trying to work at 0 V does not work as well as a standard opamp with a negative rail working at 0 volts.

I haven't looked at the specs for the A/D. Does it need to be driven from a low impedance source, or can you add a series resistor to the op-amp to limit any overvoltage current?

Richard.

[Rerouter]

First though, if you use a single LT3080, then if you want to output to go to over 30V, then you will have to reduce the current to about 250mA.

was thinking of a switching pre-regulator, biased to run approximately 5V above the output of the LT3080, so that the full 1.4A would be always available, and simply use an LC filter to cut out anything above 1KHz, though that will be slightly new to me (datasheets one is 20V)

30V+ output sounds like you have a transformer, It would be nuts if you can have a negative supply, to not use it.

i have a dual 35-0-35 transformer (68W), it would help with the current source, due to the voltage limit, and the op amps, but short of the current limit, i was thinking of ways to modify Dave's kit with minimal impact, and with options that might help others who don't have access to transformers, or not willing to buy them (laptop plugpacks spit out 24-30V)

I haven't looked at the specs for the A/D. Does it need to be driven from a low impedance source, or can you add a series resistor to the op-amp to limit any overvoltage current?

it can have 1K without any noticeable offset going by the datasheet, so i guess i could loose the clamp.

suggestions for higher quality DAC / ADC? as i was planning to write up a function to measure low resistances, (2 sense lines, accurate voltage applied and current measured with ucurrent) but the current ones are far too off to manage more than relative changes.

[jgbena]

Post here you mods/ideas for Dave's design.

Alexander.

I already saw Higher voltage, but what about 0-9 Volts, and 0-3amps instead of 1?  would that also be just a matter of changing software code and losing some resolution in the voltage and current step values?

[Rerouter]

jgbenga, the LT3080 cannot supply more than 1.4A, you would need to use its big brother the LT3083, to get 3A, being how you are only aiming at such a low voltage is would almost be a direct swap in,

[kiyotewolf]

I'd love to have it have a screen saver mode, where the LCD on bottom line, spouts Dave Jones quotes.

Also, a piezo beeper, so it can play little chirps, or warnings, or even, tunes?

A way to pick colors you want on the screen, so you can change it around later.



~Paul

[Rerouter]

the colours can be picked, so far all 3 colour channels of the lcd are wired to the micro, so pwm could be implemented, as for the screen saver thing, that would be something you might have to write in yourself, as most people want a power supply to maintain its display whether its been on 30 minutes of 100 hours,

as for the buzzer, that would probably be something you would need to mod in, for a simple buzzer, an NPN driver for a buzzer off the I/O chips,

for actual tones / alarms probably better to hook in something else running on i2c to accomplish that, as the micro will be doing a fair bit of correction if you hook it on to an active load, and you probably wouldnt want to chew up its cycles with producing a tune,

[ejeffrey]

Also, a piezo beeper, so it can play little chirps, or warnings, or even, tunes?

Are you by any chance designing hand-held instruments for agilent?!

[Rutger]

Dave, I was looking at the different options for the DAC, because I was looking at the dual version of your power supply (eg with a negative output as well) and came across the following 4 channel 12 bit DAC:
   
MCP4728-E/UN  at $ 1.44 / 100 with build in voltage reference (optional).

I know you didn't want surface mount parts but with the Maxim 4080 and the voltage ref chip you have chosen that option is already gone.

The advantage of having the extra 2 channels is that you can add/drive another power supply from the same DAC. And you can save on the cost of the parts, maybe you eliminate the external voltage ref chip.

Rutger

[caroper]

... or even, tunes?
~Paul

Who needs Overload protection when you PSU can play "Smoke on the Water, Fire in the Sky"
Joking aside, that is actually a good idea, not so much the tunes, but a beep to confirm a setting or a short alarm if CC  Mode is entered or an error condition occurs.


Cheers
Chris

[kiyotewolf]

Yeah, since the whole thing is open source, (right??), then I was thinking of modding in the little things like the tunes.. (the EEVBlog theme video end diddy, .. when it shuts off..), and the display doing EEVBlog quotes..  (as it boots, "Don't turn it on, take it apart!")

I really like having little favorite themes of colors for the RGB LCD, in case you've got mood LED lighting in your lab.. .. or for other fun reasons.  (Make it play smooth jazz chip tunes while doing little glowing of reds and oranges, like a miniature fireplace.. while powering a ... <insert thing to be powered here> )



~Paul

I was fascinated watching Dave modify the code right on the screen, and how he was able to send out serial data, twiddling the bits all by himself.

[EEVblog]

I was fascinated watching Dave modify the code right on the screen, and how he was able to send out serial data, twiddling the bits all by himself.

Which of course would have been much harder to do if I followed the advice of the "software professionals" and used a for loop instead of my "shit" code that isn't good enough for production  

Dave.

[jgbena]

jgbenga, the LT3080 cannot supply more than 1.4A, you would need to use its big brother the LT3083, to get 3A, being how you are only aiming at such a low voltage is would almost be a direct swap in,

Thanks Rerouter, ill give that a look!

[senso]

Here is my take/interpretation of Daves power supply, the pcb's have been ordered and the parts are already in Portugal.
I have maintained Dave's credit in the silkscreen and this is will be a public project in the portuguese LusoRobótica forum, it uses the smd version in the dac and the adc and also an atmega644 instead of the 328p used by Dave, it is still a beta version and I plan to shrink it a bit/a lot and make some more changes to the layout.

[george graves]

Nice work!- Now many extra pins does the 644 give you over the 328?  Are you programming it as a "sanguino?"

Looks great!

[sacherjj]

I like the break out of the sense and + separately.  Allows no burden validation of current, without the uCurrent.  (Not sure if you built that in as well.)

I'm thinking about a simpler 328 based PSU, with using the onboard ADC and PWM.  I'm currently working on a 328 based 40 V and 4 A Constant Current Sink first, but imagine the ADC and PWM experience with that will help with that.  But since I only have one decent lab PSU right now, I'll probably get this kit as well if it isn't too terribly expensive. 

Although, I hate not having the mains to the inside of the box, due to parasitic power draws or separate unplugging or power switching.  But, I might think differently if it was 220v vs 110v.

[electrode]

Wow, that looks pretty nice. Good choice of microcontroller too - 12 extra pins (answering george's question) and fully Arduino compatible.

[senso]

Thanks, I will share all the files when I'am sure that there are no errors, and yes the uCurrent is also present in this power supply.
The atmega644p as said by electrode as 12 extra pins, there will be an extra board that will serve as the front panel, for now the only displays that I have are two like this:
http://www.ebay.com/itm/0-96-128X64-OLED-LED-Display-Module-PCB-Adpater-/160556409041?pt=LH_DefaultDomain_0&hash=item2561e860d1
They are pretty tiny but can be read from pretty far and I will make a large font, and I have left access to the i2c and its possible to use any 8 bits parallel display using the SPI interface to drive some shift-registers, so it can be pretty flexible, one friend of mine wants to have one small bench supply and wants to learn how to use Eagle, so probably he will design a front panel to use with regular 16x2 char displays so I can teach in with a real project and I will make one for the small oLed.
The programming is done using AvrStudio and avr-gcc but its very easy to port to Arduino, and as the design I will put everything in my Google Code page.
If some one is interested here is my blog where I still only have some programming tutorials for avr:
http://hekilledmywire.wordpress.com/

[Zad]

If you are using SMD ICs, why not use 1206/0805 passives too? If you can solder a QFP or SOIC, you can sure as hell solder an 0805!

[Rerouter]

because a lot of hobbyists aren't keen to solder SMD's unless they really have to, and SOIC's are about as basic as SMD IC packages get,

Dave made this thinking any man with a soldering iron of no notable quality could make it himself, (or herself?) without much experience, as a kit

[electrode]

I think Zad's comment was directed at senso...

I only just got into 0805s last year. Easiest way to save board space ($$$) with footprints that "even Stevie Wonder could solder".

[senso]

I used the SOIC footprints for the ADC and DAC's because I already have then, the MAXIM parts only exist in soic or smaller, and the atmega644 in a 40 PDIP would be huge, the next iteration of the board will probably have almost all the components in SMD, and this will be first smd's that I solder, I already have some flux syringe and an JBC soldering iron with a chisel tip, wish me luck.

[electrode]

A good tip for if you don't have the gear or experience with SMD stuff is to make your pads longer. That way you can drag the excess solder back and the bridged pins will naturally separate beautifully. One of the EEVBlog vids did this.

[sacherjj]

I used the SOIC footprints for the ADC and DAC's because I already have then, the MAXIM parts only exist in soic or smaller, and the atmega644 in a 40 PDIP would be huge, the next iteration of the board will probably have almost all the components in SMD, and this will be first smd's that I solder, I already have some flux syringe and an JBC soldering iron with a chisel tip, wish me luck.

I found the cheap hot air iron to make life great with SMD.  Add some GOOD tweezers.  With that, I found it almost painless to do 0402.  Squirt on solder paste.  Add some paste flux.  Use tweezers to place component close.  Use hot air iron to melt solder paste and watch the 0402 align itself with the pads.  Pretty cool.  But if you are designing the board, use bigger components! 

[senso]

At home I have my soldering iron, flux, some fine non magnetic tweezers, fine solder and solder wick, but if everything fails in the university I can ask to use an IR soldering oven, hot air station and lots of Weller soldering stations, as I have bought the pcb's from Itead I will try first in the non E-tested boards.

[senso]

And the first pcb's finally arrived from China, 3 weeks for manufacture and shipping via slow boat, doesn't seems too bad for me, especially due to the price that I paid for them.
The silkscreen took me more time to clean than to lay out the entire pcb, and even then it has some errors/extra text that I didnt noticed.
Here are some not that good photos:

[george graves]

Looks good.  Looks like you might have a break in the ground plane - at the ground pin of ICSP header?

[senso]

It is grounded in the top ground plane, so it is in fact grounded, not a very good ground, I usually add vias to connect top and bottom ground planes.
Its funny that only the board in the photos as all the vias tented, all the others have some tented and others are not, kinda like random..

I also soldered my first SMD and it was pretty easy 
Put some flux in the pads, IC on top, hold with tweezers and a tiny amount of solder in the tip is enough for all the IC, really fast, I'm for sure converted! 

[robrenz]

Does the back and forth wander on the header holes have a purpose?  I dont know much about PCB design, just curious.

[george graves]

Ah!  I see now.

[senso]

Does the back and forth wander on the header holes have a purpose?  I dont know much about PCB design, just curious.

They are from the Sparkfun library, they call it Lock, they are off-set by 5 mills if I'm not mistaken and their purpose is that you push the headers in the holes and due to the offsets the headers get locked in placed and aligned so they are easier to solder and you dont end with crooked headers, they are really helpful when assembling the pcbs.

I dont know if Dave sees this topic, but if you are there, are you willing to share your actual code for the uSupply?

[EEVblog]

I dont know if Dave sees this topic, but if you are there, are you willing to share your actual code for the uSupply?

Here is whatever is on my machine at present.
No correspondence will be entered into 

Dave.

[senso]

Thats pretty simple for now, but well its just a power supply...
My personal version as a graphic display so I will add some current plots and who knows what more.
First board is almost complete, forgot to order some of the components and ordered the MAX4080 in the wrong package 

[electrode]

Looks really nice. The header locations look convenient too.

[senso]

One sanity check please, the uCurrent is only activated when the mosfet is turned on, and that only happens when 3.3v(high signal) is given to the 2n3904 transistor, I'm right or is it the other way around?
And sorry for not doing the homework, but what is the current threshold where I can safely enable it and not fry the MAX op-amp?

I dont know about how you will be distributing your code, but I will use a google code repository, so any-one that wants to take a sneak peek can go there, I will also offer pre-compiled hex files ready to program the atmega.

Is any one willing to try one board, I have 9, all of them e-tested and eye tested and seem to be fine, so far the only error is the 3.3v regulator must be inserted backwards ( swapped Vin with Vout  ), for some reason I used the wrong part and even with the datasheet open in front of me I didn't notice that.

I hope to have all soldered this week and at least some code doing an hello world.

[EEVblog]

FYI, I'm currently doing Rev 3, some major changes in store!
That's what happens when you leave a project for so long, you always get other ideas...
I think it's for the better though.

Dave.

[electrode]

Major == SMD?

[McMonster]

My guess is reducing BOM or going to full SMD.

[EEVblog]

Major == SMD?

SMD is part of it, the through hole kit just wasn't really going to be workable.
I'm removing some stuff and adding some other stuff!

Dave.

[EEVblog]

My guess is reducing BOM or going to full SMD.

Both, plus more.
Revising designs is half the fun

Dave.

[McMonster]

My guess is reducing BOM or going to full SMD.

Both, plus more.
Revising designs is half the fun

Dave.

If revision 3 can make coffee and order pizza then I'm getting at least two.

[mtkaalund]

If revision 3 can make coffee and order pizza then I'm getting at least two.

Would it not be easier to get a girl friend that can do tha... your right, if rev 3 can make coffee and order pizza Im also getting two

[markus_b]

I've privately toyed with a variant based on a xMega. Enough ports for everything and has 12bit AD and DA on board. Reduces the BOM an awful lot. You can program these babys with Arduino too, if you want.

The other extensions I feel useful is input voltage up to 35 Volts and current up to 3 amps (3 3080 in parallel).

[scrat]

Could a remote output voltage sensing and/or an external analog input reference be useful?
I see them as little mods, just adding one or two opamps, and using one ADC spare channel (at least in an early revision) and the uC ADC for the reference. Choosing between internal and remote sensing is usually one of the features of a precision PSU (although it could just be an hardware ORing), while the external reference could allow one to use their PSU for testing some controller (an example? A DC motor controller).

[tom66]

If the 3080s become expensive, maybe move to a power Darlington transistor or a P-channel MOSFET with a different driver circuit? Disadvantage - no current sense or temperature sense on die, but first current sense is done by the 317T and temperature sense could be done with an external thermistor and NPN to pull the drive signal low as an emergency cut off.

Opened up my Tagasaki power supply I got a while ago. 0-35V 0-3A. Quite interesting construction - the circuit is very similar to Dave's but no micro of course. Just two single op-amp packages, I think they were similar to 741's, and a few passive carbon resistors. Some 10W resistors x 3 for current sense. And a TL431 for voltage reference. What I found really clever though was how they were choosing taps on the transformer to reduce power dissipation in the 3 x 2N3055's (on a massive heatsink.) They used some kind of SCR devices I think but I never quite figured out how they got them to fire - I'm almost certain they weren't using any logic, probably just biasing the gate somehow so it fired when the output voltage was past a certain point...? The transformer in the power supply was easily the heaviest part, and the power supply is only capable of 105W output max.

Nowadays with modern laptop SMPS's going spare you can easily find a 20V DC power supply so there's no need to use a big bulky transformer. You could even get the laptop SMPS to follow the output by a couple of volts to significantly reduce power dissipation. There's probably a way to tap into the feedback to do this. I've done it before with an ATX supply but only increased voltage (12V -> 26V) not decreased it.

[george graves]

I guess I'm starting to worry about the price!  (not too much - it looks like this kit is going to be a must-have, and a great bang for the buck!)

Will SMD increase the cost?  Decrease it?  I'd assume that you'd be doing a lot less labor with SMD as you wouldn't have to sort out all the parts and put them into a bag.

[EEVblog]

I guess I'm starting to worry about the price!  (not too much - it looks like this kit is going to be a must-have, and a great bang for the buck!)
Will SMD increase the cost?  Decrease it?  I'd assume that you'd be doing a lot less labor with SMD as you wouldn't have to sort out all the parts and put them into a bag.

Yes, SMD will be cheaper and less hassle for all involved, albeit at the expense of losing the satisfaction of having soldered it yourself for some people.

Dave.

[Zad]

In this weeks' The Amp Hour, Dave points out that the regulator chip is now available in a switched mode equivalent.

Just sayin'....

[EEVblog]

In this weeks' The Amp Hour, Dave points out that the regulator chip is now available in a switched mode equivalent.

Close, but no cigar.

I think I'll shoot a new video tomorrow revealing all.
Rev 3 is coming along nicely and I'm quite pleased with this version. I've eliminated some chips, lowered some costs, increased some functionality, and improved the build at the same time.

Dave.

[dougg]

Just tried to hack up a USB power supply from a 5 volt, 2 amp switched mode wall slug. With USB D+ and D- floating it didn't work on two different tablets (HP and RIM). Then I followed Dave's power supply schematic which puts 2 volts on both with separate resistor dividers. That did not work either. The appropraite reference to look up is "USB dedicated charger". Seems that the USB folks standardized shorting D+ and D- together circa 2010. When I did that, it worked.

[EEVblog]

Just tried to hack up a USB power supply from a 5 volt, 2 amp switched mode wall slug. With USB D+ and D- floating it didn't work on two different tablets (HP and RIM). Then I followed Dave's power supply schematic which puts 2 volts on both with separate resistor dividers. That did not work either. The appropraite reference to look up is "USB dedicated charger". Seems that the USB folks standardized shorting D+ and D- together circa 2010. When I did that, it worked.

The iPhone/Pad/Pod is apparently different and requires those voltages.

Dave.

[Monkeh]

Just tried to hack up a USB power supply from a 5 volt, 2 amp switched mode wall slug. With USB D+ and D- floating it didn't work on two different tablets (HP and RIM). Then I followed Dave's power supply schematic which puts 2 volts on both with separate resistor dividers. That did not work either. The appropraite reference to look up is "USB dedicated charger". Seems that the USB folks standardized shorting D+ and D- together circa 2010. When I did that, it worked.

The iPhone/Pad/Pod is apparently different and requires those voltages.

Dave.

Yes, Apple aren't big on this whole standards thing.

[sacherjj]

I thought the D+ D- tie was a change that allowed 1 A through USB for charging, rather than just 500 mA.  If those are not shorted, you are only allowed to pull 500 mA. 

[mtkaalund]

I thought the D+ D- tie was a change that allowed 1 A through USB for charging, rather than just 500 mA.  If those are not shorted, you are only allowed to pull 500 mA.

I believe that the specification for the USB standard, says that 500mA is the maximum allowed current through an USB port. But correct me if I am wrong

[Hideki]

500 mA was correct for a long time, but as time went by various useful things were added to the specification.

See http://www.usb.org/developers/devclass_docs for the Battery Charging Specification documents.

[Monkeh]

I thought the D+ D- tie was a change that allowed 1 A through USB for charging, rather than just 500 mA.  If those are not shorted, you are only allowed to pull 500 mA.

I believe that the specification for the USB standard, says that 500mA is the maximum allowed current through an USB port. But correct me if I am wrong

500mA is the maximum current a USB 2.0 port can supply to a device, unless it complies with the USB Battery Charging Specification. Shorting D+ and D- is required to indicate a dedicated charging port, which can supply up to 5A.

USB 3.0 includes support for up to 900mA at reduced voltages (making allowance for cheap, nasty cables, basically).

[dougg]

I thought the D+ D- tie was a change that allowed 1 A through USB for charging, rather than just 500 mA.  If those are not shorted, you are only allowed to pull 500 mA.

I believe that the specification for the USB standard, says that 500mA is the maximum allowed current through an USB port. But correct me if I am wrong

"The great thing about standards are that there are so many to pick from." You were correct but as one standard supersedes another you are probably not correct now. I am familiar with SCSI standards and recent USB-3 external disks should be using UASP rather than the (dreadful) mass storage class (block mode). I'm still looking for the first example in a real product.

[EEVblog]

Part 11 is rendering now, the biggest and most rambling episode yet, at 1:15min! 

The true original purpose of the PSU will be revealed, along with the substantial changes.

Damn, 1:15min, I must be insane. Now I'm thinking if there is somehow a way to split it?...

Dave.

[markus_b]

Dave, just keep it like it is. People always can fast-forward !

[firewalker]

I am guessing 4xNiMh batteries and a boost convert.

Alexander.

[Tooms]

Part 11 is rendering now, the biggest and most rambling episode yet, at 1:15min! 

The true original purpose of the PSU will be revealed, along with the substantial changes.

Damn, 1:15min, I must be insane. Now I'm thinking if there is somehow a way to split it?...

Dave.

love them as they are, so dont split it.

Tooms

[sacherjj]

Am I the only one that checked about a dozen times today, each time when I got back to my desk, for a new 1:15 monster video on YouTube?

[Noize]

Am I the only one that checked about a dozen times today, each time when I got back to my desk, for a new 1:15 monster video on YouTube?

No 

[PeterG]

Its getting on  for midday. I am thinking something may have gone horribly wrong with the video causing delays.

Regards

[IanB]

Yup, it's been hours and it hasn't appeared yet. I think Dave broke YouTube 

[BravoV]

Nah .. after waking up this morning, Dave decided to scap and redo new version again, and going to re-shoot the vid ...

....j/k.

Seriously, I think he hasn't upload it yet and probably he was rendering it overnight.

[EEVblog]

I decided to try splitting into two, a hardware/case part, and the schematic part.
Plus I had to baby sit Sagan for half the day!

Dave.

[mtkaalund]

I decided to try splitting into two, a hardware/case part, and the schematic part.
Plus I had to baby sit Sagan for half the day!

Dave.

When can we expect the schematic part? I was looking forward to see it, but i like the system design aspect of this video.
Sagan, an electronic engineer in the making?

[EEVblog]

It STILL doesn't have Bluetooth!!?

It's got a socket with SPI interface, plug in anything you want!

Dave.

[firewalker]

Or a serial one.

http://iteadstudio.com/store/index.php?main_page=product_info&cPath=18&products_id=179

[PeterG]

To my thinking, Bluetooth makes more sense than Ethernet. I think Ethernet defeats the purpose of being battery powered and portable.

I would have liked to see the accuracy increase rather than decrease. It would be great to have 1mV and 1mA accuracy but i feel that is outside the scope of this project for now.

But thats just my 2 cents.

Regards

[McMonster]

To my thinking, Bluetooth makes more sense than Ethernet. I think Ethernet defeats the purpose of being battery powered and portable.

I would have liked to see the accuracy increase rather than decrease. It would be great to have 1mV and 1mA accuracy but i feel that is outside the scope of this project for now.

But thats just my 2 cents.

Regards

I agree, it'd be nice to figure out a way to either plug a Bluetooth or Ethernet module there.

[markus_b]

I agree, it'd be nice to figure out a way to either plug a Bluetooth or Ethernet module there.

The Bluetooth module is easy, you can use the Arduino-compatible serial connector to connect  it. A Bluetooth module needs Gnd, 3.3V, RX and TX. These are all there.

[McMonster]

I agree, it'd be nice to figure out a way to either plug a Bluetooth or Ethernet module there.

The Bluetooth module is easy, you can use the Arduino-compatible serial connector to connect  it. A Bluetooth module needs Gnd, 3.3V, RX and TX. These are all there.

My thought was "something that fits there without some hacking or bodgy wires". The socket would be most likely placed so it aligns to the hole in the case and there's probably no direct replacement BT module that would match the socket pins. So we're either mount the module somewhere else and do some wiring or design a custom daughter board that can be plugged nicely, fit's the space and is mounted securely for moving the whole unit. Not that someone would play american football with the PSU, but if it's portable then the construction must be prepared for thing you don't do with most of the bench equipment. Or at least it must be built like old school bench hardware, I'm pretty sure I can airdrop my Wavetek without a chute and it'll survive.

[sacherjj]

I agree, it'd be nice to figure out a way to either plug a Bluetooth or Ethernet module there.

The Bluetooth module is easy, you can use the Arduino-compatible serial connector to connect  it. A Bluetooth module needs Gnd, 3.3V, RX and TX. These are all there.

My thought was "something that fits there without some hacking or bodgy wires". The socket would be most likely placed so it aligns to the hole in the case and there's probably no direct replacement BT module that would match the socket pins. So we're either mount the module somewhere else and do some wiring or design a custom daughter board that can be plugged nicely, fit's the space and is mounted securely for moving the whole unit. Not that someone would play american football with the PSU, but if it's portable then the construction must be prepared for thing you don't do with most of the bench equipment. Or at least it must be built like old school bench hardware, I'm pretty sure I can airdrop my Wavetek without a chute and it'll survive.

I'm betting there would be space to have an adapter board, with pins and headers.